Skip to Main Content
The detection and tracking of people is an important aspect in the video analysis and a popular topic in computer vision as well. This paper mainly discusses the detection and tracking of people in video sequences captured from a stationary camera. An effective people detection algorithm is proposed based on the bi-directional projection histogram of grayscale two-frame differencing image, proposing an effective method for multiple people segmentation. A simple applied method, called directional nearest neighbor matching, is developed for tracking people. Large numbers of experimental results on real video sequences from outdoor and indoor scenes have demonstrated the real-time performance (25~30 frames/s), a high accuracy for sparse crowd and robustness for environments change.